cloudscale-slb-controller

A kubernetes controller that watches for services of type LoadBalancer and creates a floating IP on cloudscale.ch for those services. The floating IP will be attached to the server where one of the services pods in state Running is located.

It is safe to run multiple replicas of this controller.

supported service annotations

The following annotations on services of type LoadBalancer are supported:

• linkyard.ch/slb-controller-id: this service must be processed by a controller with the given id
• linkyard.ch/existing-floating-ip: use the already existing floating IP on cloudscale.ch for this service; note that no additional checks (e.g. port collisions) are performed
• linkyard.ch/preserve-floating-ip: don't delete the floating IP on cloudscale.ch when the service is deleted (or otherwise changed in a way that the IP no longer belongs to it). Default is false

configuration

The following configuration flags are supported:

• -cloudscale-token, CLOUDSCALE_TOKEN, no default, required: cloudscale.ch API token with write access
• -ip-limit, IP_LIMIT, default 5: maximum number of floating ips that are allowed to exist on the cloudscale.ch account; this is used as an additional guard in case this controller is out of control and would create an infinite amount of floating ips
• -controller-id, CONTROLLER_ID, no default: if set, the controller will only process services with a linkyard.ch/slb-controller-id annotation set to this value
• -leader-election-configmap, LEADER_ELECTION_CONFIGMAP, no default: name of the ConfigMap to use for leader election
• -leader-election-namespace LEADER_ELECTION_NAMESPACE, no default: name of the namespace where the ConfigMap for leader election is located at
• -leader-election-node, LEADER_ELECTION_NODE, no default: name of the pod
• -leader-election-ttl, LEADER_ELECTION_TTL, default 10s: TTL for leader election, e.g. 10s; valid time units are ns, us, ms, s, m and h.
• -log-level, LOG_LEVEL, default Debug: log level to set; possible values are: Panic, Fatal, Error, Warn, Info, Debug, Trace
• -kubeconfig, KUBECONFIG, no default, not required: path to a kubernetes kubconfig file; this is not required when running inside a kubernetes cluster
• -fake-cloudscale-client, FAKE_CLOUDSCALE_CLIENT: set to true to use a fake cloudscale.ch API client
• -chaos-chance, CHAOS_CHANCE, default 0: chance of a call to one of the fake components failing; range [0.0,1,0)